SSH Command - NAT Acceleration for AI Mesh Node

[abracadabra11]

Is there an SSH command to look up whether NAT Acceleration is active on an AI Mesh Node?

Recently connected by AI Mesh Node via MOCA to the Main Router and I'm seeing about 500Mbps when conducting iperf3 testing from my wired Desktop (ethernet connection to AI Mesh Node) to the Main Router. When I remove the Node and connect the Desktop directly to the MOCA adapter, throughput jumps to about 750Mbps so there appears to be a bottleneck with the Node (RT-AC68U).

 

[ColinTaylor]

I think the best clue would be to run top on the node and then see if the CPU load goes to 100% when you do your speed test.

 

[abracadabra11]

I think the best clue would be to run top on the node and then see if the CPU load goes to 100% when you do your speed test.

I did that earlier and I was hitting around 75% or so. Not 100% so I suspected that this wasn't the limiting factor, but wanted to check command directly if available.

 

[ColinTaylor]

NAT (HW) acceleration is a kernel model that accelerates traffic as it is routed between the WAN and the LAN. Assuming the node works much like an access point it doesn't have a WAN interface and is not routing anything so the NAT acceleration is not applicable.

 

[abracadabra11]

NAT (HW) acceleration is a kernel model that accelerates traffic as it is routed between the WAN and the LAN. Assuming the node works much like an access point it doesn't have a WAN interface and is not routing anything so the NAT acceleration is not applicable.

That's good to know, but the AI Mesh Node is connected through the WAN port to the Main Router. So perhaps it's still relevant?

 

[RMerlin]

No. The nodes don't do any NAT, only the primary router does. So hardware acceleration is only relevant to the main router.

 

[ColinTaylor]

Your post did remind me of a "feature" I came across when trying to debug an anomaly with QoS using iperf in John's fork a few years ago.

The relevant part here was that with my RT-AC68U the maximum router to LAN throughput I could get was about 600 Mbps. Here's part of a conversation I had at that time:

The discrepancy does appear to be caused by generic-segmentation-offload being set on br0. Unfortunately whilst turning off GSO does correct the numbers it also significantly reduces the Router to LAN throughput. The maximum throughput with GSO is about 600Mbps, without it it drops to half that . In both cases the throughput is limited by the router's CPU (and TCP buffer size).

It's a shame that TC can't seem to work this out, especially since in different circumstances it could really mess up the traffic flows. Maybe it's a bug. Ironically the TLDP documentation even mentions GSO and recommends turning it off to improve latency.

"Since TSO, GSO, UFO and GRO greatly increase the number of bytes which can be queued in the driver queue you should disable these optimizations if you want to optimize for latency over throughput. It’s doubtful you will notice any CPU impact or throughput decrease when disabling these features unless the system is handling very high data rates."

That last sentence is doubly ironic.

 

[abracadabra11]

Your post did remind me of a "feature" I came across when trying to debug an anomaly with QoS using iperf in John's fork a few years ago.

The relevant part here was that with my RT-AC68U the maximum router to LAN throughput I could get was about 600 Mbps. Here's part of a conversation I had at that time:

That's interesting - thanks.

I was thinking about reconfiguring the node in AP mode and retesting that way, but I've got a lot of cable ties securing the wiring for both devices (among other things) and am hesitant to test and find that I should go back to the AI Mesh configuration (i.e. if similar throughput, might as well just keep AI Mesh for configuration simplicity). Moving the AI Mesh node to rebind it to the main router would not be fun.

Any idea if there are SSH commands that I can use to backup the current AI Mesh node configuration so that I can just flash the configuration rather than pairing the devices again?

 

[ColinTaylor]

Any idea if there are SSH commands that I can use to backup the current AI Mesh node configuration so that I can just flash the configuration rather than pairing the devices again?

I don't use AiMesh but I would assume it's the usual NVRAM "Save settings" normally done through the GUI. From the command line that's:

nvram save yourfilename